OHDSI in Korea
Rae Woong Park, MD, PhD
Department of Biomedical Informatics,
Ajou University, School of Medicine,
Korea
Contents
Korean OHDSI Network
Why DRN and CDM is popular to Korea?
Lesson learned from potential Data Owner
Future plans in Korean OHDSI
2
Korean OHDSI network: 18
목동
마곡
CDM Conversion Status
Conversion completed: 4 institutions
Ajou University Hospital: 2.3M, 23 years EHR
Gachun Gil University Hospital: 2M, 10 years EHR
Kangwon National University Hospital: 0.5M, 10 years EHR
NHIS: 2M, 12 years, Claim + regular health exam (2018: 51M, 12 years)
Conversion in progress 14 institutions
By the end of 2017: 2
Samsung Medical Center
Wonkwang University Hospital
Wonju Severance Hospital
By the end of 2018: 14
Chonbuk National University Hospital
Yonsei Unviersity Severance Hospital
Korea University Anam Hosital
Korea University Guro Hospital
Korea University Ansan Hospital
Gangdong Sacred Heart Hospital
Hanyang University Hospital
Ehwa Unviersity Mokdong Hospital
Ehwa Unviersity Magok Hospital
Cha University Hospital
HIRA/NHIS (51M, 12 years)
<MOU>
Activity in Korea
Leadership meeting
Bimonthly
Leaders in charge of CDM for
each hospital
Important decisions and policy-
related issues
Engineer meeting
Biweekly, TC (current 9
th
)
EHR experts from participating
hospitals
Discuss all the technical issues
during CDM conversion
5
Activity in Korea
Open forum
Monthly, 3 hour lecture
Agenda
Introduction to OHDSI and CDM
OMOP CDM Structure
OMOP CDM Vocabulary/ vocabulary
mapping
Tools for OMOP CDM
ETL process
Research Experience using OHDSI
Network
6
CDM conversion Standard Operation Procedure (SOP)
Step Time-line
Human Resources
From Data Partner
Role of Data
Partner
Role of Supporting
Organization
Contact
1
MOU
CDM director
and
engineer
Administrative affair
Administrative affair
shinda@ajou.ac.kr
2
Advance
Preparation
-
CDM director
and
engineer
Budget for
H/W, S/W,
human resources
Orientation, lecture and
consultation
3
Vocabulary
Extraction
1
wk
EMR expert
Frequency
table of
every drug, supply,
procedure, diagnosis,
lab test
Support
hidoyebi@ajou.ac.kr
4
Vocabulary
Mapping
-
Vocabulary
customization: 2 mon
- Code mapping: 3 mon
Medical expert x 2
(physician x1, Nurse x 1)
Support
Code mapping between
local code and OMOP
vocabulary (
voca
covering
99% of data)
5
ETL
Definition
2 mon
EMR expert
x1
Medical expert x 1
ETL document
Support
enzo@ajou.ac.kr
bacojun@ajou.ac.kr
6
ETL
2
- 4 weeks
DB expert x 1
ETL Query
Support
7
OHDSI
Tool
application
1
- 4 weeks
R/JAVA
expert x 1
Achilles, Atlas,
etc
installation
Support
jsh90612@gmail.com
8
DQM
4
-8 weeks
DB expert
Medical
expert
Run DQM codes
Provide DQM codes
and evaluation
and
feedback
enzo@ajou.ac.kr
bacojun@ajou.ac.kr
1) SOP for MOU
8
2) SOP for Advance Preparation
9
3) SOP for Vocabulary Extraction
10
4) SOP for Vocabulary Mapping
11
5) SOP for ETL documentation
Vojtech Huser, MD, PhD
12
GOVERNMENT’S INTEREST ON CDM
13
Korean government formally launched a task-force
team for distributed bio-dig data sharing
14
Minister of Mistry
of Trade, Industry
and Energy of
South Korea
formally
announced that
they launch a TF
team for
distributed bio-dig
data sharing to
build a national
biomedical data
sharing platform
and environment.
This year, they will
make a master
plan for it and
prepare budget to
realize it.
Apr 18, 2017
15
Hospit
al
Bio Big Data
Center
IT company
Compa
ny
Distributed Bio Big Data Model
Analysis Results
Analysis SW
Analysis request
Analysis Results
Analysis SW
supply
Analysis SW
request
NHIS*, Development of a CDM-based Drug Safety
Surveillance System
16
Three year project
2016-208
1
st
year: feasibility
12 year of 1M pt data into
CDM
2
nd
year; validation of
usefulness of the CDM
3
rd
year : Full conversion
12-years of 51 M patients
*National Health Insurance Service: Governmental National Health Insurer
OMOP CDM conversion
RAW data
SAS file?
ETL Layer
Voca
mapper
ETL
Data
Conversio
n
DQM
Analyses
Evaluation
Desing
Conversio
n
OMOP CDM
Conversion
Standard Vocabularies
RxNor
m
LOINC
SNOMED-
CT
Application of pharmacovigilance tools
PV tools
ROR, PRR
LGPS
CLEAR
CERT
OHDSI tools
ATLAS ACHILLES
Claim data,
NHIS
Collaboration and competition in between government
Ministry of Trade, Industry and Energy
Ministry of Health and Welfare
National Health Insurance Service
Ministry of Food And Drug Safety
Ministry of Science, ICT and future planning
DRN, OMOP-CDM
EMR + Omics + life-log
merge, HL7 CDA?
OMOP-CDM?
EMR + Claim
DRN, OMOP-CDM
Claim + Health Exam
DRN, K-CDM (?)
EMR
HIS with CDM
Characteristics of Korean OHDSI
Korean OHDSI
Data partners:
Major tertiary teaching hospitals
Detailed time stamp
Test results
Outcome data
National Health Insurance Data
Compulsory health insurance
Claim data + socioeconomic data + regular Health exam data
(includes lab tests)
12-year of observation period
Covers all the citizens (51M)
WHY DRN AND CDM IS POPULAR TO
KOREA?
19
Invited
Talks
157 invited talks during
past 33 months since
July 2014.
2014: 13 times
2015: 42 times
2016: 74 times
2017: 28 times
No. of Presentations and Data Partner Join
21
157
76
12 ln(presentation)
18 months
0
2
4
6
8
10
12
14
Jun-14
Jul-14
Aug-14
Sep-14
Oct-14
Nov-14
Dec-14
Jan-15
Feb-15
Mar-15
Apr-15
May-15
Jun-15
Jul-15
Aug-15
Sep-15
Oct-15
Nov-15
Dec-15
Jan-16
Feb-16
Mar-16
Apr-16
May-16
Jun-16
Jul-16
Aug-16
Sep-16
Oct-16
Nov-16
Dec-16
Jan-17
Feb-17
Mar-17
Apr-17
May-17
N. of Presntaition
(Log scale)
Date
Presentation Join
Gacheon
NHIS
Lesson
learned from
potential
Data Owner
Quick-prototyping
Live demonstration
Success story
Focusing on clinicians
Lesson
learned from
potential
Data Owner
Quick-prototyping
Launch Achilles ASAP!
Need vocabulary mapping!
International
Standard
Korean Standard Code
SNOMED CT
ICD-10-PCS
CPT-4
ATC
RxNorm
LOINC
Dx
Drug
Surgery
Lab
Pathology
Anesthesia
Radiology
EDI
EDI
EDI
EDI
EDI
ICD9-CM
KCD-6
24
Code Mapping
>170K codes to map
EDI (Electronic Data Interchange) code in Korea
National standard for national health insurance claim
Managed by Health Insurance Review & Assessment Service (HIRA)
Strategy for Standard Vocabulary Mapping
Sort codes by frequency of usage
Number of codes required to cover
95 % of transaction data 99% 100%
20-30% of codes 60-70% of codes
25
No. of drug codes (Ajou Univ.)
Total number of
prescriptions
(27%)
DRUG code mapping (Ajou Univ.)
26
Source code: local code of Ajou university hospital (mapped with
EDI, partially)
Standard code: RxNorm, ATC
64,360,565
2,228
28,616,311
988
23,102,063
736
415,125
1,240
55,357
41
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
No. of prescriptions
No. of drugs
RxNorm clinical drug RxNorm ingredient ATC No EDI code No match
24.4% (1,281/5,233)75.64% (3,952/5,233)
99.6%
55.2%
24.6%
19.8%
61.5%
14.1%%
Lesson
learned from
potential
Data Owner
Live demonstration using
RWD
(Distributed research network)
Achilles: data characterization
Openness: http://ami.ajou.ac.kr:8080
29
Malignant tumor, Breast
Ajou Univ. Hospital 2.3M 22years
Malignant tumor, Breast
NHISS, 1M sample cohort, 10 years
Lesson
learned from
potential
Data Owner
Using a Success story and
role model
Examples of clinical researches using big data
31
250M patients
12 database
5 countries
Lesson
learned from
potential
Data Owner
Focusing on Clinicians
Focusing on clinicians
Clinicians, rather than informaticians: major decision
makers in a hospital are usually clinicians.
Young clinicians (assistant associate professors) :
highest interest on OMOP CDM, because they do not
have enough fund, resources and data for their
research.
Inter-disciplinary clinical meeting
then homogenous clinical meeting at initial stage
33
Invited Talks by domain
34
0
20
40
60
80
2014 2015 2016 2017
Clinical Govermental IT etc
ETC
IT meeting
Governmental meeting
Clinical meeting
Number of
presentations
28
13
42
74
35
HOW ABOUT PROVIDE FUND OR
INCENTIVES TO HOSPITAL?
Not successful
Challenges and plans in Korean OHDSI
Expanding the data partners
All the tertiary teaching hospitals
Most of general hospitals
Some of private clinics
Most of pharmacies
All the claim data covering all the
Korean population
Governance for data sharing
No IRB for pre-defined and
verified analyses
Open all the Achilles website of
Korean data partner to public
Maintain regular leadership
meeting
Free sample data
For training
To test analytic code
For Feasibility test
Real-time CDM
Up to date information
Real-time eligibility screening
CDSS
CDM-based PHR
Expansion of CDM model for
bio-signal data
Genomic data
Image data
36
Summary
About South Korea
Korean OHDSI Network
Why DRN and CDM is popular to Korea?
Lesson learned from potential Data Owner
Quick-prototyping
Live demonstration
Success story
Focusing on clinicians
Projects/Tools
Future plans in Korean OHDSI
37
Acknowledgements
Dukyong Yoon, MD, PhD
Hyun Wook Han, MD, PhD
Song Vogue Ahn, MD, PhD
Soo Yeon Cho, MPH
DaHye Shin, BS
JungHyun Byun, BS
Hojun Park, BS
MinSeok Jeon, BE
Sungjae Jung, BE
Doyeop Kim, BE
Sanghyung Jin, MS
Tae Young Kim, BE
Seojeong Shin, MS
Jaehyeong Cho, BS
Lee hye jin, MBA
Eun Sung Kim BS
Jaehee Han, BN
So Young Eo, BN
Hyo Hung Kim,BN
Eugene Jeong, BS
Seungbin Oh, BPharm
Yourim Lee, BS
Seng Chan You, MD, MS
Haze Lee
Jang ho Lee, BE
Taehwan Kim, BE
A Rum Lee, BS
EunBi Kim, BS
Yeo Kyung Lee, BS
Seong-yun Oh
Hyukjun Cho, MS
EunKang Kim
Yun Yeon Ju
Yujung Kwon
Soojung Cho, MPH
38